For vendors
How we evaluate products
If you build infrastructure for AI workloads, here is exactly how we assess it and how to get in front of us. There is no fee, and there is no way to pay for placement.
What this is, and what it is not
We are an independent advisory. Engineering teams come to us with a problem, and we recommend the architecture and the products that fit it. We are paid by them, not by you.
That means there is no listing package to buy, no sponsored placement, and no way to influence a ranking. If we ever have a referral arrangement with you, we disclose it to the client on the recommendation itself.
What we can offer is more useful than a listing anyway: we talk to engineering teams actively trying to solve the problem your product solves, and we would rather describe your product accurately than vaguely.
What we ask about
The same questions for every product, so that two of them can genuinely be compared.
The problem it removes
The specific technical failure or operational bottleneck it addresses — not the marketing category. Many products in this space describe themselves identically and behave very differently.
Fit and scale
The workload type and scale where it works well, and where it stops working. A tool that suits a shared cluster of fifty accelerators is often wrong for five, and wrong again for five thousand. Whether it assumes a dedicated platform team matters just as much.
Deployment and operational overhead
Self-hosted, managed, or Kubernetes-dependent. Who operates it once it is in, and how it behaves at three in the morning.
Unstated prerequisites
The most common reason a good product fails is an assumption nobody mentioned — a platform team, an existing cluster, a particular network fabric.
Commercial model and exit cost
Open source, usage-based, or committed contract. What it costs to leave matters as much as what it costs to start.
Weaknesses and failure modes
Every product has a shape. We write the trade-off down, because a recommendation without one is not a recommendation.
The one thing we will not do
We will not remove the trade-off. Every product on this site has a "what to watch out for" section, including yours. If your product genuinely does not suit a particular situation, we will say so — that is the entire reason engineering teams trust the rest of it.